[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20221114193940.whiwpojrwast7tnl@soft-dev3-1>
Date: Mon, 14 Nov 2022 20:39:40 +0100
From: Horatiu Vultur <horatiu.vultur@...rochip.com>
To: Alexander Lobakin <alexandr.lobakin@...el.com>
CC: <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
<bpf@...r.kernel.org>, <davem@...emloft.net>,
<edumazet@...gle.com>, <kuba@...nel.org>, <pabeni@...hat.com>,
<ast@...nel.org>, <daniel@...earbox.net>, <hawk@...nel.org>,
<john.fastabend@...il.com>, <UNGLinuxDriver@...rochip.com>
Subject: Re: [PATCH net-next 1/5] net: lan966x: Add XDP_PACKET_HEADROOM
The 11/14/2022 15:16, Alexander Lobakin wrote:
> [Some people who received this message don't often get email from alexandr.lobakin@...el.com. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
>
> From: Horatiu Vultur <horatiu.vultur@...rochip.com>
> Date: Sun, 13 Nov 2022 12:15:55 +0100
Hi Olek,
>
> > Update the page_pool params to allocate XDP_PACKET_HEADROOM space as
> > headroom for all received frames.
> > This is needed for when the XDP_TX and XDP_REDIRECT are implemented.
> >
> > Signed-off-by: Horatiu Vultur <horatiu.vultur@...rochip.com>
> > ---
>
> [...]
>
> > @@ -466,7 +472,8 @@ static struct sk_buff *lan966x_fdma_rx_get_frame(struct lan966x_rx *rx,
> >
> > skb_mark_for_recycle(skb);
> >
> > - skb_put(skb, FDMA_DCB_STATUS_BLOCKL(db->status));
> > + skb_put(skb, FDMA_DCB_STATUS_BLOCKL(db->status) + XDP_PACKET_HEADROOM);
> > + skb_pull(skb, XDP_PACKET_HEADROOM);
>
> These two must be:
>
> + skb_reserve(skb, XDP_PACKET_HEADROOM);
> skb_put(skb, FDMA_DCB_STATUS_BLOCKL(db->status));
>
> i.e. you only need to add reserve, and that's it. It's not only
> faster, but also moves ::tail, which is essential.
Thanks for the suggestion. I will do that in the next version.
>
> >
> > lan966x_ifh_get_timestamp(skb->data, ×tamp);
> >
>
> [...]
>
> > --- a/drivers/net/ethernet/microchip/lan966x/lan966x_xdp.c
> > +++ b/drivers/net/ethernet/microchip/lan966x/lan966x_xdp.c
> > @@ -44,8 +44,9 @@ int lan966x_xdp_run(struct lan966x_port *port, struct page *page, u32 data_len)
> >
> > xdp_init_buff(&xdp, PAGE_SIZE << lan966x->rx.page_order,
> > &port->xdp_rxq);
> > - xdp_prepare_buff(&xdp, page_address(page), IFH_LEN_BYTES,
> > - data_len - IFH_LEN_BYTES, false);
> > + xdp_prepare_buff(&xdp, page_address(page),
> > + IFH_LEN_BYTES + XDP_PACKET_HEADROOM,
> > + data_len - IFH_LEN_BYTES - XDP_PACKET_HEADROOM, false);
>
> Are you sure you need to substract %XDP_PACKET_HEADROOM from
> @data_len? I think @data_len is the frame length, so headroom is not
> included?
Actually @data_len contains also the XDP_PACKET_HEADROOM. I am calling
lan966x_xdp_run like this:
---
return lan966x_xdp_run(port, page,
FDMA_DCB_STATUS_BLOCKL(db->status) + XDP_PACKET_HEADROOM);
---
But from what I can see, that doesn't make too much sense.
Because I am adding here the XDP_PACKET_HEADROOM and then remove it in
xdp_prepare_buff. So it would be better not to add it at all.
>
> > act = bpf_prog_run_xdp(xdp_prog, &xdp);
> > switch (act) {
> > case XDP_PASS:
> > --
> > 2.38.0
>
> Thanks,
> Olek
--
/Horatiu
Powered by blists - more mailing lists